Skip to content

Wire Briefings Search

Fetching from the wire…

01

02

03

04

05

06

07

08

Wire Briefings Search

Entity trail

KAT-Coder-V2

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

0

Findings

1

Edges

4

Sources

1

Corpus findings

2026-04-01 / arxiv-researcherKAT-Coder-V2: Agentic Coding Model Hits 79.6% SWE-bench via Specialize-then-Unify TrainingKuaishou's KAT-Coder-V2 decomposes agentic coding into five expert domains (SWE, WebCoding, Terminal, WebSearch, General), trains each independently with SFT and RL, then consolidates via on-policy distillation into a single model. Achieves 79.6% on SWE-bench Verified (vs Claude Opus 4.6 at 80.8%) and 88.7 on PinchBench surpassing GLM-5 and MiniMax M2.7. Built on KwaiEnv infrastructure sustaining tens of thousands of concurrent sandbox instances, with a novel Tree Training technique achieving 6.2x speedup for tree-structured RL trajectories.

Graph relationships

BENCHMARKED AGAINST
Claude Opus 4.6
KAT-Coder-V2 compared to Claude Opus 4.6 performance
Source finding
BENCHMARKED AGAINST
SWE-bench Verified
KAT-Coder-V2 achieved 79.6% on SWE-bench Verified
Source finding
RELEASED
Kuaishou
Kuaishou released KAT-Coder-V2
Source finding
RELEASED
Kuaishou -> KAT-Coder-V2
Kuaishou released KAT-Coder-V2
Source finding

Source trail

Graph sources

entity graphfindings textkg edgeskg entities

KAT-Coder-V2 intelligence trail | MindPattern